PCA-Driven Multivariate Trait Integration in Alfalfa Breeding: A Selection Model for High-Yield and Stable Progenies

Breeding improvement in alfalfa (Medicago sativa L.) is often constrained by the complexity of agronomic traits and trade-offs among yield-related characteristics. Conventional single-trait selection rarely captures the full range of phenotypic variation or the interactions among traits. To address this, we developed a principal component analysis (PCA)-based framework for multivariate selection in hybrid breeding. Six yield-related traits—plant height, branch number, fresh/hay yield ratio (FHR), leaf/stem ratio (LSR), multifoliolate leaf frequency, and dry weight per plant—were quantified in two parental lines and their F1/F2 generations. PCA identified three principal components (PC1–PC3) with eigenvalues >1, explaining 71.14% of the total phenotypic variance: PC1 (32.43% variance) was predominantly loaded with positive contributions from dry weight per single plant, height, and branches, biologically representing overall plant vigor and biomass accumulation; PC2 (21.77% variance) showed strong negative loadings for LSR, capturing architectural trade-offs between stem dominance and leaf production; PC3 (16.94% variance) had positive loadings on multifoliolate leaf rate and fresh/dry ratio, embodying quality and physiological resilience traits. Based on PCA scores, a composite selection index was constructed, and the top 31.1% of F1 hybrids were selected. Their F2 progeny showed significant improvements in dry weight (+15.56%, p < 0.01), multifoliolate leaf frequency (+74.78%, p < 0.001), and reduced FHR (–8.2%, p < 0.05), accompanied by lower yield decline (−7.2% versus −14.1% in controls). These results show that PCA-based multivariate selection effectively balances trait trade-offs, enhances intergenerational stability, and improves selection efficiency. This framework offers a practical tool for alfalfa breeding.